def __init__(self,
spark,
experimentID,
model_name,
input_path,
limit=None):
self.spark = spark
self.input_path = input_path
self.model_name = model_name
self.limit = limit
self.experimentID = experimentID
self.data_provider = LendingClubDataProvider(spark, input_path, limit)
def __init__(self,
spark,
input_path,
output_path,
model_name,
stage,
limit=None):
self.spark = spark
self.input_path = input_path
self.output_path = output_path
self.model_name = model_name
self.limit = limit
self.stage = stage
self.data_provider = LendingClubDataProvider(spark, input_path, limit)
class LendingClubConsumerPipeline():
def __init__(self,
spark,
input_path,
output_path,
model_name,
stage,
limit=None):
self.spark = spark
self.input_path = input_path
self.output_path = output_path
self.model_name = model_name
self.limit = limit
self.stage = stage
self.data_provider = LendingClubDataProvider(spark, input_path, limit)
def run(self):
df = self.data_provider.load_and_transform_data_consumer()
df.createOrReplaceTempView("loans")
self.spark.udf.register(
"model",
mlflow.pyfunc.spark_udf(self.spark,
model_uri='models:/{}/{}'.format(
self.model_name, self.stage)))
self.spark.sql("""
select *, model(term, home_ownership, purpose, addr_state, verification_status, application_type, loan_amnt, emp_length,
annual_inc,dti, delinq_2yrs, revol_util, total_acc, credit_length_in_years, int_rate, net, issue_year) as prediction
from loans
""").write.format('delta').mode('overwrite').save(self.output_path)
class LendingClubTrainingPipeline():
def __init__(self, spark, input_path, model_name, limit=None):
self.spark = spark
self.input_path = input_path
self.model_name = model_name
self.limit = limit
self.data_provider = LendingClubDataProvider(spark, input_path, limit)
def run(self):
X_train, X_test, Y_train, Y_test = self.data_provider.run()
self.train(X_train, X_test, Y_train, Y_test)
def train(self, X_train, X_test, Y_train, Y_test):
cl = LogisticRegression(random_state=42, max_iter=100)
# cl = RandomForestClassifier(random_state=42)
cl.fit(X_train, Y_train)
with mlflow.start_run(run_name="Training") as run:
self.eval_and_log_metrics(cl, X_test, Y_test)
mlflow.sklearn.log_model(cl, "model")
def eval_and_log_metrics(self, estimator, X, Y):
predictions = estimator.predict(X)
# Calc metrics
acc = accuracy_score(Y, predictions)
roc = roc_auc_score(Y, predictions)
mse = mean_squared_error(Y, predictions)
mae = mean_absolute_error(Y, predictions)
r2 = r2_score(Y, predictions)
# Print metrics
print(" acc: {}".format(acc))
print(" roc: {}".format(roc))
print(" mse: {}".format(mse))
print(" mae: {}".format(mae))
print(" R2: {}".format(r2))
# Log metrics
mlflow.log_metric("acc", acc)
mlflow.log_metric("roc", roc)
mlflow.log_metric("mse", mse)
mlflow.log_metric("mae", mae)
mlflow.log_metric("r2", r2)
mlflow.set_tag('candidate', 'true')
class LendingClubModelEvaluationPipeline():
def __init__(self,
spark,
experimentID,
model_name,
input_path,
limit=None):
self.spark = spark
self.input_path = input_path
self.model_name = model_name
self.limit = limit
self.experimentID = experimentID
self.data_provider = LendingClubDataProvider(spark, input_path, limit)
def run(self):
mlflow_client = MlflowClient()
_, X_test, _, Y_test = self.data_provider.run()
cand_run_ids = self.get_candidate_models()
best_cand_roc, best_cand_run_id = self.get_best_model(
cand_run_ids, X_test, Y_test)
print('Best ROC (candidate models): ', best_cand_roc)
try:
versions = mlflow_client.get_latest_versions(self.model_name,
stages=['Production'])
prod_run_ids = [v.run_id for v in versions]
best_prod_roc, best_prod_run_id = self.get_best_model(
prod_run_ids, X_test, Y_test)
except RestException:
best_prod_roc = -1
print('ROC (production models): ', best_prod_roc)
if best_cand_roc >= best_prod_roc:
# deploy new model
model_version = mlflow.register_model(
"runs:/" + best_cand_run_id + "/model", self.model_name)
time.sleep(5)
mlflow_client.transition_model_version_stage(
name=self.model_name,
version=model_version.version,
stage="Production")
print('Deployed version: ', model_version.version)
# remove candidate tags
for run_id in cand_run_ids:
mlflow_client.set_tag(run_id, 'candidate', 'false')
def get_best_model(self, run_ids, X, Y):
best_roc = -1
best_run_id = None
for run_id in run_ids:
roc = self.evaluate_model(run_id, X, Y)
if roc > best_roc:
best_roc = roc
best_run_id = run_id
return best_roc, best_run_id
def get_candidate_models(self):
spark_df = self.spark.read.format("mlflow-experiment").load(
self.experimentID)
pdf = spark_df.where("tags.candidate='true'").select(
"run_id").toPandas()
return pdf['run_id'].values
def evaluate_model(self, run_id, X, Y):
model = mlflow.sklearn.load_model('runs:/{}/model'.format(run_id))
predictions = model.predict(X)
# acc = accuracy_score(Y, predictions)
roc = roc_auc_score(Y, predictions)
return roc